← All guides

Editing, explained

Manuscript analysis: what gets measured, and what it can and can't tell you

"Manuscript analysis" is the diagnostic step of editing: examining a whole book to find out what is working, what is not, and where — before anyone changes a word. It can be done by a human editor reading closely, by software measuring the text, or both. The honest version of this guide is that each kind of analysis is good at different things, and authors waste money when they buy one expecting the other. Here is what actually gets measured, what the numbers mean, and what no analysis can tell you.

The four layers a full analysis covers

LayerWhat is examinedBest analysed by
StructureWhere the hook, inciting incident, midpoint shift, dark moment and climax fall; whether the ending pays off the opening; subplot trackingA whole-book reader — human or an AI that ingests the full manuscript
PacingScene and chapter length patterns, the ratio of scene to summary, dialogue-to-narrative balance, where tension rises and dropsMeasured by software, interpreted by judgement
Character and arcGoal, motivation, conflict per protagonist; whether change is earned; consistency of behaviourClose reading; see what an editor checks in a character arc
ProseLexical variety, word and phrase repetition, filter words, adverb density, sentence-length variation, passive constructions, dialogue-tag patternsSoftware — these are countable, and humans count them badly

What the prose metrics actually mean

Software analysis produces numbers. The numbers are only useful if you know what a good one looks like, so here are the common ones and how to read them:

  • Lexical variety — the share of distinct words in the text. Low variety reads as flat; artificially high reads as thesaurus-driven. The useful comparison is against published books in your genre, not an abstract ideal.
  • Repetition — words and phrases that recur more than the genre baseline. Character names and "said" are supposed to repeat; "a moment later" and "let out a breath" are not.
  • Filter words — saw, felt, heard, noticed, realised, wondered. Each one puts the narrator between the reader and the experience. A high count is the single most fixable prose problem in a first draft.
  • Adverb density — not "delete every -ly" but how many are propping up weak verbs. The rate matters more than the total.
  • Sentence-length variation — monotone rhythm is invisible on the page and obvious in the numbers.
  • Dialogue ratio — romance commonly runs dialogue-heavy; a long stretch with almost none is often where the pace dies.

The test of a good software analysis is repeatability. If the same draft gets a different score on a second run, the number cannot measure progress. Authors Starport's Instant Score runs nine deterministic craft metrics across the whole manuscript — lexical variety, repetition, filter words, adverb density and more — benchmarked against professionally edited books, and the same book always scores the same, so a change in the score is a change in the book.

Human vs software analysis — the honest split

Software is better at counting. It will find every filter word in 90,000 words in seconds, and it will not get tired in chapter 23. A human is better at meaning: whether the heroine's refusal in chapter 12 is in character, whether the twist is earned, whether the voice is distinctive enough to build a readership on. Tools like ProWritingAid, AutoCrit and Grammarly do the counting layer well for sentences and paragraphs; the book editing software guide compares them. What most of them cannot do is the structural layer, because they analyse the text in front of them rather than the book as a whole.

Whole-manuscript AI analysis now sits in the middle. It reads the entire book, so it can report structure — where the hook really starts, where the middle sags, which arc disappears — and it can do it on every draft. Authors Starport's Developmental Critique is that read for romance: verdict, ranked problem areas and chapter notes, each anchored to a passage in your text. Scene Map lays the structure out scene by scene so the pacing analysis is something you can see.

What no analysis can tell you

  • Whether the book is good. Analysis finds problems; it does not measure delight.
  • Whether a rule-break is working. A metric flags the 40-word sentence; only a reader knows it was the right 40 words.
  • Whether the market wants it. That is a different question with different tools — Market Fit for the romance shelf, and honest comp-title research.
  • Whether you should publish. Analysis gets you to a strong draft; the decision is yours.

Use the analysis for what it is: the fastest, cheapest way to find out which kind of editing the book needs and to prove each revision moved the needle. The free Reader-Ready Scorecard is the place to start; what a manuscript report contains shows what the human-style write-up looks like.

Find out where your manuscript actually stands — free.

The Reader-Ready Scorecard grades your full manuscript on nine craft metrics, benchmarks it against professionally-edited books, and tells you exactly what to fix first. Free, in minutes.

Get your free Reader-Ready Scorecard →

FAQ

What is manuscript analysis?

Manuscript analysis is the diagnostic examination of a whole book — its structure, pacing, character arcs and prose — to identify what is working and what is not before any editing begins. It can be done by a human editor reading closely, by software measuring the text, or by AI tools that ingest the full manuscript and do both.

What does manuscript analysis software measure?

Typically lexical variety, word and phrase repetition, filter words, adverb density, sentence-length variation, passive constructions, dialogue-to-narrative ratio and dialogue-tag patterns. Whole-manuscript tools also report structural features such as where the hook and midpoint fall and how scene length varies across the book.

Is AI manuscript analysis accurate?

For countable prose features, software is more accurate than a human and far faster; the important property is repeatability — the same draft should score the same every time. For structural and character judgements, whole-manuscript AI analysis is useful and increasingly reliable, but its notes should be read as a strong editor's opinion to test against the book, not a verdict.

How is manuscript analysis different from a manuscript evaluation?

Analysis is the examination; an evaluation is a human editor's service that delivers the resulting report. In practice, 'manuscript analysis' usually refers to the measurable, software-assisted layer and 'manuscript evaluation' to the editorial judgement layer. A complete diagnostic uses both.

How much does manuscript analysis cost?

A human manuscript evaluation commonly runs $600–$1,500 for a full-length novel. Software analysis ranges from free to subscription tools costing a few hundred dollars a year. Authors Starport's Reader-Ready Scorecard analyses the full manuscript on nine craft metrics free; deeper reads such as the Developmental Critique are priced on the plans page.

More guides